If I was starting a new project from zero today, this is exactly where I’d start.
It doesn’t matter if the project is a business, a family project, research, content, software, or something completely different.
The foundation can stay the same.
THE FOUNDATION
AGENTS.md tells the agents how to operate.
PROJECT.md defines what we’re actually trying to accomplish.
STATUS.md keeps everyone grounded in where the project stands right now.
DECISIONS.md preserves the important choices so they don’t disappear into old conversations.
THE ORGANIZATION
inbox/ gives everything new a temporary landing place.
areas/ breaks the project into whatever major sections naturally make sense for that specific project.
resources/ holds the files, references, assets, notes, data, and other material the project depends on.
THE WORK
work/ separates what’s queued, what’s actively being worked on, and what’s already completed.
outputs/ gives finished or review-ready work a clear destination.
archive/ keeps old or replaced material without cluttering the active project.
THE SCALING PART
This is where it gets useful.
The root AGENTS.md contains the rules that apply to the entire project.
If one area becomes complicated enough, that folder can get its own AGENTS.md with instructions specific to that area.
If it grows again, you can repeat the same pattern another level down.
The goal isn’t to build a massive system on day one.
Start with a universal foundation.
Then specialize downward only when the complexity actually earns it.
STANFORD JUST LEAKED A FREE LECTURE ON HOW CHATGPT AND CLAUDE ACTUALLY WORK.
1 hour 40 minutes. free.
Bookmark this and send it to a friend.
2.8 million views, and most people using AI every day still haven't watched it.
ChatGPT → How It Works → Better Prompts → Real Leverage
then read my guide on which model to actually run once you get it, GPT-6 Astra vs Kimi K3, below ↓